Verdict Schema

FUD.ai returns a structured verdict with exactly 8 fields. This schema is registered on the CROO Dashboard as the Service DeliverableType — any mismatch will cause on-chain settlement to fail.

Field reference

Field	Type	Description
`executable_verdict`	enum	Action directive: `ACCUMULATE`, `LIQUIDATE_LONGS`, `HOLD`, `IGNORE_FUD`, `INSUFFICIENT_DATA`
`drama_index`	number	Composite intensity index (0-100). Combines chatter level and risk score.
`confidence`	number	Pipeline confidence in the verdict (0.0-1.0). `0` when degraded.
`dominant_branch`	string	The winning scenario from MCTS analysis (e.g. `false_fud`, `real_crash`).
`evidence_chain`	string[]	Human-readable evidence statements supporting the verdict.
`coordination_signals`	object	Sybil & coordination metrics from social data.
`served_from_cache`	boolean	Whether this result was served from ingestion cache.
`branch_probabilities`	Record<string, number>	Probability distribution across all MCTS branches.

executable_verdict

The primary action directive. Your agent should branch on this field.

Value	Meaning	Color	Action
`ACCUMULATE`	FUD is false, price likely to recover	Green	Consider buying
`LIQUIDATE_LONGS`	Real crash detected	Red	Exit long positions
`IGNORE_FUD`	Noise without actionable signal	Green	Hold / ignore
`HOLD`	Insufficient conviction either way	Yellow	Maintain current position
`INSUFFICIENT_DATA`	Pipeline degraded, data incomplete	Gray	Do not act

drama_index

A composite score from 0-100 that measures the intensity of the situation:

drama_index = round(0.4 * chatter_level + 0.6 * risk_score)

0-39: Low intensity — normal market noise
40-69: Moderate — worth monitoring
70-100: High — significant event in progress

confidence

How confident the pipeline is in the verdict, from 0.0 to 1.0. This is calibrated by the reflexion loop based on historical prediction accuracy for similar market conditions.

> 0.7: High confidence — safe to act on
0.4-0.7: Moderate — consider additional signals
< 0.4: Low — treat with caution
0: Pipeline was degraded, verdict is not usable

coordination_signals

This is FUD.ai's core differentiator. It tells you whether the social panic is organic or manufactured.

Field	Type	Interpretation
`unique_author_ratio`	number (0-1)	< 0.3 = highly coordinated botnet, > 0.6 = organic
`duplicate_text_cluster_size`	number	> 5 = copy-paste bot campaign, ≤ 2 = low duplication
`cross_platform_burst_window_minutes`	number	< 10 min = synchronized burst, > 60 min = spread organically

branch_probabilities

The MCTS engine evaluates three parallel scenarios. The full probability distribution is exposed so your agent can hedge:

{
  "real_crash": 0.12,
  "false_fud": 0.65,
  "whale_manipulation": 0.23
}

The dominant_branch is simply the key with the highest probability.

Example verdict

{
  "executable_verdict": "ACCUMULATE",
  "drama_index": 38,
  "confidence": 0.82,
  "dominant_branch": "false_fud",
  "evidence_chain": [
    "[SECURITY] No honeypot detected, contract is verified",
    "[SYBIL] unique_author_ratio: 0.71 — organic discussion pattern",
    "[ON-CHAIN] No abnormal exchange inflows detected"
  ],
  "coordination_signals": {
    "unique_author_ratio": 0.71,
    "duplicate_text_cluster_size": 2,
    "cross_platform_burst_window_minutes": 0
  },
  "served_from_cache": false,
  "branch_probabilities": {
    "real_crash": 0.12,
    "false_fud": 0.65,
    "whale_manipulation": 0.23
  }
}